The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute the codex CLI tool using a Bash shell to perform analysis and get coding advice.
[COMMAND_EXECUTION]: The instructions explicitly guide the agent on how to grant the external service write access to the workspace using the -s workspace-write flag, which allows an external entity to modify local files.
[DATA_EXFILTRATION]: To function, the skill sends repository context, file contents, and specific prompts to an external service (OpenAI's Codex via the CLI).
[DATA_EXFILTRATION]: The skill recommends writing the output of the external service to /tmp/codex-consult.txt, potentially exposing sensitive technical analysis in a shared temporary directory.
[PROMPT_INJECTION]: The skill is designed to be 'self-triggering,' meaning the agent is encouraged to decide on its own when to invoke this external process, which could be used to bypass user oversight for specific tasks or decision-making processes.
[COMMAND_EXECUTION]: The mechanism for passing the agent's prompt to the shell command ("<your prompt>") could be susceptible to command injection if the agent includes un-sanitized content or special shell characters in its query.

consult-codex